9 Charging Sites: What Failed After Commissioning

Most coverage of EV charging infrastructure is about the build: permits, capital, grid applications, ribbon cuttings. Very little is written about the year after handover, when the site is supposedly working and the operator starts paying for reality.

We went back to nine sites we had commissioned across our project portfolio — 100+ customers in 22 countries — and asked a deliberately unglamorous question: what actually failed in the first year? Not what was slow to install. Not what the utility delayed. What broke, degraded, or quietly stopped delivering, after the site was signed off.

The answer was consistent, and it was rarely the chargers.

The equipment mostly held

Across the nine sites, the DC power modules and dispensers accounted for very little of the downtime. That matters because procurement conversations still revolve around which hardware to buy. On the evidence of these sites, hardware was not the weak link.

What failed was the electrical design around the hardware — the decisions made before a single charger was ordered.

Transformer headroom came first

Several sites had been designed against the nameplate capability of the equipment rather than the load the utility would actually approve. The gap only becomes visible under real traffic, when the site cannot deliver the advertised power at peak. Fixing it afterwards is not a spare-parts problem; it is a renegotiation with the utility, a new protection study, and in some cases a second application cycle. The operators who avoided this had asked a simple question early: what load will the utility approve, on paper, today?

Protection coordination was the second recurring failure

Settings carried over from another project, or left at defaults, produced nuisance tripping that looked to drivers like a broken charger. One site tripped on inrush at the start of the evening peak for three weeks before anyone traced it. The repair was inexpensive; the diagnosis was not, because the fault appeared intermittently and only under load.

Cooling and installation position came third

Equipment specified for a temperature band, then installed against a wall or in afternoon sun, derates exactly when demand is highest. This is the least discussed failure mode because it does not produce error codes — the charger keeps working, just slower. Operators often discover it only when revenue per connector does not match the model.

A fourth pattern appeared in the numbers rather than the hardware: metering drift

Two sites showed a gap between energy delivered at the connector and energy billed, small per transaction and significant per month. Nobody had set a reconciliation routine, so the discrepancy was discovered at the quarterly review rather than at commissioning. The fix was a reference meter and a two-point check, which takes an afternoon — but only if someone owns the process.

What this means for the operator, not the engineer

The commercial consequence of these EV charging station failures is rarely a dramatic outage. It is a site that delivers 70–80 percent of its modelled revenue for a year, and a business case that quietly stops matching reality. That is harder to detect than a dead charger, and much harder to explain to an investor or a lender.

For site hosts — hotels, fuel stations, commercial landlords — the implication is narrower but useful. The host controls very little of the electrical design, but the host does control site access, the position of the equipment, and the timing of handover. Those three levers determine whether the commissioning window is two weeks or two months.

For fleet operators, the lesson is about redundancy. A depot that depends on a single transformer has no fallback when that transformer is derated or under review. The question worth asking before signature is not “how fast does it charge” but “what happens on the day the site cannot deliver full power”.

And one organisational failure that cost more than any of the above

On three of the nine sites the protection settings actually applied on site did not match the settings in the document that was handed over. Nothing failed because of it during the review period, but every future maintenance visit would have started from the wrong baseline. A handover pack is not paperwork; it is the operating manual for the next ten years.

Three lessons worth carrying into the next site

First, commission against the real grid, not an assumed one. Pre-commissioning on a temporary supply proves the equipment, the communications and the payment stack. It cannot prove the site against the fault level and capacity it will actually see. That test belongs before handover, not after.

Second, write the EV charging commissioning criteria into the contract, with numbers. “Works as intended” is not a criterion. “Delivers the contracted power per connector at the specified ambient temperature, with protection settings per the approved sheet, evidenced by a signed test record” is a criterion. Sites where this was agreed in advance did not have a handover argument.

Third, treat the first year as a designed phase, not a warranty period. Operators who planned a structured six-month review — capacity versus actual demand, protection events, temperature derating, failure modes by unit — caught most of these issues while remediation was still cheap. Operators who waited for driver complaints found out at the worst possible time, during peak season, with the asset already in service.

None of this is an argument against building. EV charging infrastructure is still one of the more defensible infrastructure investments in emerging markets, where the constraint is usually capacity rather than demand. It is an argument for spending the design phase on the parts that do not appear in a product brochure.

The nine sites are still running. The ones that were designed around the grid they had — rather than the grid they hoped for — are the ones that have not needed to go back to the utility.